Expand my Community achievements bar.

Announcement: Calling all learners and mentors! Applications are now open for the Adobe Analytics 2024 Mentorship Program! Come learn from the best to prepare for an official certification in Adobe Analytics.
SOLVED

Adobe Analytics - Attribution Models: 1 Unique Order ID got multiple Orders for Attribution Models other than Last Touch

Avatar

Level 1

Hi, 

I'm trying out the different Attribution Models within Adobe Analytics using the Attribution Panel. Something very strange happened.

 

It's a Marketing Channel Report with Orders as the metric:

  • Lockback Window: Visitor (Reporting Window)
  • I broke down the Affiliate Channel by Order ID. 
  • The First Touch () column, shows 4 first touches for one Order ID.
Spoiler
  • Bildschirm­foto 2023-11-20 um 11.51.32.png

There is a similar thread already, but since the solution way is not provided. I thought I should raise a new thread. 

 

How can I investigate, if it is a duplicate problem in my case? Or the issue is rooted somewhere else? 

 

Thanks in advance 

 

 

Topics

Topics help categorize Community content and increase your ability to discover relevant content.

1 Accepted Solution

Avatar

Correct answer by
Level 1

That is not the Issue. 

The issue appears, using Dimension, which does not make sense in combination with Attribution Models.

The Logic behind the different Attribution Models is always the same, regardless of which dimension. That means First Touch will attribute the metric (here order) to the very first value of the selected Dimension.

 

In my case, First Touch has 4 Orders for this one Order ID, since this one was the First Order ID value appearing within the Customer Journey. And this Person did 4 Orders in Total. So that's why the same Order ID has different counts for different Attribution Models.  

View solution in original post

5 Replies

Avatar

Community Advisor

When it comes to testing a scenario, working within an isolated environment (like dev) and being able to segment down to your specific user are important.

 

Then you can run through various purchases, keeping track of the order ids for each and which order you made those purchases in...

 

Then you can look at your own isolated user with the models knowing exactly what transactions transpired and when, and start to drill deeper into what the attributions are showing you.

Avatar

Correct answer by
Level 1

That is not the Issue. 

The issue appears, using Dimension, which does not make sense in combination with Attribution Models.

The Logic behind the different Attribution Models is always the same, regardless of which dimension. That means First Touch will attribute the metric (here order) to the very first value of the selected Dimension.

 

In my case, First Touch has 4 Orders for this one Order ID, since this one was the First Order ID value appearing within the Customer Journey. And this Person did 4 Orders in Total. So that's why the same Order ID has different counts for different Attribution Models.  

Avatar

Community Advisor

For Order ID "1234"

Last Touch|Visitor = on the Order metric, this would be the actual order "1234"...  Orders after 1234 would then get a different "Last Touch" (i.e. 1235 overwrites 1234, then 1236 overwrites 1235, etc)

 

First Touch|Visitor = on the Order metric is the FIRST order recorded for the Marketing Channel Period, you had 4 orders in that period, therefore, the first order id recorded would be shown... ie. 1234

 

Linear|Visitor = this splits the attribution credit equally across the Visitor... so I guess in this case, you had 4 purchases, therefore order 1234 is given a percentage of the attribution.... (I think it might be a 50% attribution.. so 50% of "4" is "2")

 

To be perfectly honest, I wouldn't have used Attribution Modeling against Order like this... Marketing Channel is already an attribution model... I don't understand why your would then apply a secondary attribution model on top of it... 

 

 

My suggestion to run an isolated test with specific known Order IDs would allow you to look at each of the 4 order ids knowing the exact path that was taken and see what the attribution was doing to ALL of them...  Hands on experience helps to build understanding about what is happening.

 

 

As it stands, I don't see anything wrong with the attribution calculations per se (aside from the usage itself not really making a lot of sense - as I said, I don't understand why you would apply an attribution model to something like Marketing Channel, when it already is an attributed value...)

Avatar

Level 1

Thanks for the further explanation. I would love to elaborate on that. 

Linear|Visitor = this splits the attribution credit equally across the Visitor... so I guess in this case, you had 4 purchases, therefore order 1234 is given a percentage of the attribution.... (I think it might be a 50% attribution.. so 50% of "4" is "2")


Yes, that makes total sense, and we are on the same page on that. 

 

To be perfectly honest, I wouldn't have used Attribution Modeling against Order like this... Marketing Channel is already an attribution model... I don't understand why your would then apply a secondary attribution model on top of it... 

 


I wanted to understand the Linear Model better since in the provided Overview sheet it just says, that is evenly attributed but not if the order value gets divided by the occurring Dimension ( 3 Channels, every Channel gets 1/3 of the Order or, every Channel gets 1 ). 

Arvindto_0-1701415689245.png

Source

That's why I thought, looking at 1 specific OrderID would explain it to me.

 

Now that I know, how the Models behave, and it is reasonable for me, I have no further questions, as of now ;). 

 

BR and Thanks again

 

Avatar

Community Advisor

Honestly, Attribution is hard to understand sometimes...  I think that maybe the linear model works better for something like Revenue... against your Marketing Channels...

 

In the last touch model, the revenue belongs to the most recent driver... but it's also possible that other Marketing Efforts helped lead to the sale, therefore the linear model allows you to split the revenue across the touch points. So some of the money made was due to Search, and some of the money was due to a social media campaign, and some of the money was due to a Marketing email, etc....